Julie Carreau

Timeline

1. 2004

   Locally linear embedding for QSAR

   In 2004, she coauthors a paper with Yoshua Bengio, Olivier Delalleau and S. Y. Yue in the Journal of Computer-Aided Molecular Design on locally linear embedding for dimensionality reduction in QSAR.

2. 2007–2009

   Hybrid Pareto models with Bengio

   She completes a B.Sc. in pure mathematics, an M.Sc. in computational and mathematical finance, and a Ph.D. in computer science at the Université de Montréal, where she defends her doctoral thesis on hybrid Pareto models for asymmetric heavy-tailed distributions in December 2007. With Yoshua Bengio, she presents a hybrid Pareto model for conditional density estimation—replacing the upper tail of a Gaussian distribution with a generalized Pareto tail—at the Eleventh International Conference on Artificial Intelligence and Statistics in 2007. Their paper on a hybrid Pareto mixture for conditional asymmetric fat-tailed distributions appears in IEEE Transactions on Neural Networks in 2009.

3. November 2009–May 2010

   Postdoctoral fellowships in France

   From November 2009 to May 2010, she is a UJF postdoctoral fellow in the MISTIS project at Inria near Grenoble, one of two postdoctoral fellowships she holds in France along with one at the Laboratoire des Sciences du Climat et de l'Environnement near Paris.

4. 2011

   Downscaling precipitation at IRD Montpellier

   After her postdoctoral fellowships in France, she holds a researcher position, junior then senior, for 11 years at the Institut de recherche pour le développement in Montpellier, spending the last two years in Tunis at the École Supérieure des Communications de Tunis. In 2011, she and Mathieu Vrac publish a paper in Water Resources Research on stochastic downscaling of precipitation with neural network conditional mixture models.

5. June 2021

   Assistant professor at Polytechnique Montréal

   In June 2021, she joins Polytechnique Montréal as an assistant professor in the Department of Mathematical and Industrial Engineering, a role she holds until May 2026. A 2021 paper with Vincent Guinot in Advances in Water Resources presents a PCA spatial pattern based artificial neural network downscaling model for urban flood hazard assessment.

6. 2022–2023

   Flood wave propagation and IVADO

   Since 2022, she serves as an associate editor of Stochastic Environmental Research and Risk Assessment, and her term as an IVADO researcher runs from 2023 to 2028. In 2023, she and Philippe Naveau publish a spatially adaptive multi-resolution generative algorithm for simulating flood wave propagation in Weather and Climate Extremes, which extends the discrete wavelet lifting scheme to spatio-temporal data and which she presents at a GERAD seminar in June 2023.

7. 2025–2026

   Probabilistic U-Net and climate downscaling

   In 2025, she receives a Springer Nature Author Service Award and coauthors a probabilistic U-Net approach to downscaling climate simulations that combines a deterministic U-Net backbone with a variational latent space to capture aleatoric uncertainty, presented as a poster at the AI for Science workshop at NeurIPS 2025 in San Diego in December 2025. She also supervises master's theses on deep learning for climate downscaling and for precipitation dated 2025 and 2026, and in 2026 coauthors Spatial Pattern Regression to reconstruct gridded meteorological fields from regional climate model simulations and station observations.

8. Since June 2026

   Associate professor and Mila member

   In June 2026, she is promoted to associate professor in the Department of Mathematics and Industrial Engineering at Polytechnique Montréal. She is an associate academic member of Mila, with deep learning and representation learning as her research topics.